LU-1285 tests: fix tar breakage in 102d/102f/102j

[fs/lustre-release.git] / lustre / tests / replay-ost-single.sh

diff --git a/lustre/tests/replay-ost-single.sh b/lustre/tests/replay-ost-single.sh
index 5c8d9d3..7727f38 100755 (executable)
--- a/lustre/tests/replay-ost-single.sh
+++ b/lustre/tests/replay-ost-single.sh
@@ -2,25 +2,20 @@
 
 set -e
 
-PTLDEBUG=${PTLDEBUG:--1}
 LUSTRE=${LUSTRE:-$(cd $(dirname $0)/..; echo $PWD)}
 SETUP=${SETUP:-""}
 CLEANUP=${CLEANUP:-""}
 . $LUSTRE/tests/test-framework.sh
 init_test_env $@
 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
+init_logging
 
 # While we do not use OSTCOUNT=1 setup anymore,
 # ost1failover_HOST is used
 #ostfailover_HOST=${ostfailover_HOST:-$ost_HOST}
 #failover= must be defined in OST_MKFS_OPTIONS if ostfailover_HOST != ost_HOST
 
-remote_ost_nodsh && skip "remote OST with nodsh" && exit 0
-
-if [ "$FAILURE_MODE" = "HARD" ] && mixed_ost_devs; then
-    skip "$0: Several ost services on one ost node are used with FAILURE_MODE=$FAILURE_MODE. "
-    exit 0
-fi
+require_dsh_ost || exit 0
 
 # Tests that fail on uml
 CPU=`awk '/model/ {print $4}' /proc/cpuinfo`
@@ -107,24 +102,68 @@ test_4() {
 }
 run_test 4 "Fail OST during read, with verification"
 
+iozone_bg () {
+    local args=$@
+
+    local tmppipe=$TMP/${TESTSUITE}.${TESTNAME}.pipe
+    mkfifo $tmppipe
+
+    echo "+ iozone $args"
+    iozone $args > $tmppipe &
+
+    local pid=$!
+
+    echo "tmppipe=$tmppipe"
+    echo iozone pid=$pid
+
+    # iozone exit code is 0 even if iozone is not completed
+    # need to check iozone output  on "complete"
+    local iozonelog=$TMP/${TESTSUITE}.iozone.log
+    rm -f $iozonelog
+    cat $tmppipe | while read line ; do
+        echo "$line"
+        echo "$line" >>$iozonelog
+    done;
+
+    local rc=0
+    wait $pid
+    rc=$?
+    if ! $(tail -1 $iozonelog | grep -q complete); then
+        echo iozone failed!
+        rc=1
+    fi
+    rm -f $tmppipe
+    rm -f $iozonelog
+    return $rc
+}
+
 test_5() {
-    [ -z "`which iozone 2> /dev/null`" ] && skip "iozone missing" && return 0
-    FREE=`df -P $TDIR | tail -n 1 | awk '{ print $4/2 }'`
-    GB=1048576  # 1048576KB == 1GB
-    if (( FREE > GB )); then
-        FREE=$GB
+    [ -z "`which iozone 2> /dev/null`" ] && skip_env "iozone missing" && return 0
+
+    # striping is -c 1, get min of available
+    local minavail=$(lctl get_param -n osc.*[oO][sS][cC][-_]*.kbytesavail | sort -n | head -1)
+    local size=$(( minavail * 3/4 ))
+    local GB=1048576  # 1048576KB == 1GB
+
+    if (( size > GB )); then
+        size=$GB
     fi
-    IOZONE_OPTS="-i 0 -i 1 -i 2 -+d -r 4 -s $FREE"
-    iozone $IOZONE_OPTS -f $TDIR/$tfile &
-    PID=$!
-    
+    local iozone_opts="-i 0 -i 1 -i 2 -+d -r 4 -s $size -f $TDIR/$tfile"
+
+    iozone_bg $iozone_opts &
+    local pid=$!
+
+    echo iozone bg pid=$pid
+
     sleep 8
     fail ost1
-    wait $PID
-    RC=$?
-    log "iozone rc=$RC"
+    local rc=0
+    wait $pid
+    rc=$?
+    log "iozone rc=$rc"
     rm -f $TDIR/$tfile
-    [ $RC -ne 0 ] && return $RC || true
+    [ $rc -eq 0 ] || error "iozone failed"
+    return $rc
 }
 run_test 5 "Fail OST during iozone"
 
@@ -137,27 +176,39 @@ test_6() {
 
     f=$TDIR/$tfile
     rm -f $f
-    sync && sleep 2 && sync    # wait for delete thread
+    sync && sleep 2 && sync  # wait for delete thread
+
+    # wait till space is returned, following
+    # (( $before > $after_dd)) test counting on that
+    wait_mds_ost_sync || return 4
+    wait_destroy_complete || return 5
+
     before=`kbytesfree`
     dd if=/dev/urandom bs=4096 count=1280 of=$f || return 28
     lfs getstripe $f
+    stripe_index=$(lfs getstripe -i $f)
+
     sync
-    sleep 2                                    # ensure we have a fresh statfs
+    sleep 2 # ensure we have a fresh statfs
     sync
-#define OBD_FAIL_MDS_REINT_NET_REP       0x119
-    do_facet mds "lctl set_param fail_loc=0x80000119"
+
+    #define OBD_FAIL_MDS_REINT_NET_REP       0x119
+    do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000119"
     after_dd=`kbytesfree`
     log "before: $before after_dd: $after_dd"
     (( $before > $after_dd )) || return 1
     rm -f $f
-    fail ost1
+    fail ost$((stripe_index + 1))
+    wait_recovery_complete ost$((stripe_index + 1)) ||
+        error "OST$((stripe_index + 1)) recovery not completed"
     $CHECKSTAT -t file $f && return 2 || true
     sync
     # let the delete happen
-    sleep 5
+    wait_mds_ost_sync || return 4
+    wait_destroy_complete || return 5
     after=`kbytesfree`
     log "before: $before after: $after"
-    (( $before <= $after + 40 )) || return 3   # take OST logs into account
+    (( $before <= $after + 40 )) || return 3   # take OST logs into account
 }
 run_test 6 "Fail OST before obd_destroy"
 
@@ -165,10 +216,16 @@ test_7() {
     f=$TDIR/$tfile
     rm -f $f
     sync && sleep 5 && sync    # wait for delete thread
+
+    # wait till space is returned, following
+    # (( $before > $after_dd)) test counting on that
+    wait_mds_ost_sync || return 4
+    wait_destroy_complete || return 5
+
     before=`kbytesfree`
     dd if=/dev/urandom bs=4096 count=1280 of=$f || return 4
     sync
-    sleep 2                                    # ensure we have a fresh statfs
+    sleep 2 # ensure we have a fresh statfs
     sync
     after_dd=`kbytesfree`
     log "before: $before after_dd: $after_dd"
@@ -176,16 +233,18 @@ test_7() {
     replay_barrier ost1
     rm -f $f
     fail ost1
+    wait_recovery_complete ost1 || error "OST recovery not done"
     $CHECKSTAT -t file $f && return 2 || true
     sync
     # let the delete happen
-    sleep 5
+    wait_mds_ost_sync || return 4
+    wait_destroy_complete || return 5
     after=`kbytesfree`
     log "before: $before after: $after"
     (( $before <= $after + 40 )) || return 3   # take OST logs into account
 }
 run_test 7 "Fail OST before obd_destroy"
 
-equals_msg `basename $0`: test complete, cleaning up
+complete $(basename $0) $SECONDS
 check_and_cleanup_lustre
-[ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG && grep -q FAIL $TESTSUITELOG && exit 1 || true
+exit_status